Re: All Graphics Cards related queries here.
DISCLAIMER:
Please do not take these as writing in stone. Just my logic and reasoning. Open for debate...as always.
Many times we debate the CPU: GPU ratio, primarily because the CPU feeds the data to the GPU. A CPU bottle neck (in this post CPU bottle neck is referred to bottle necking the GPU, no other computer device) in layman terms can be described as "the CPU is holding back the GPU in terms of processing. The GPU has to 'wait' for the CPU to send it more data to process." So I decided to run some benchmarks tests to verify this.
Also if people find this logical, they can use the same method(s) to check their CPU bottleneck.
Methodology:
1. Installed Rivatuner (for load monitoring).
2. Patched .CFG file to realize the HD4890 GPU.
3. Started the Hardware Monitoring on Rivatuner -- let system idle for 15-20 seconds.
4. Ran specific 3D stress utility.
5. Upon exit of (4) at times automatic or manual again let system idle for 15-20 seconds.
6. Made screenshot of Rivatuner.
System Specifications:
CPU : E7400 (Wolfdale - 3M)
FSB X Multiplier : 381 Mhz x 10.5 = 4.00 Ghz
Motherboard : MSI Platinum
Chipset : P45 Intel Eaglelake/Intel 82801JR ICH10R
Bios Version : AMI v1.06 (02/20/2009)
Ram : Corsair XMS2 CM2X2048-8500C5 4 GB DDR2-800 DDR2 @ 1016 Mhz.
Latency : 5-5-5-15
Graphic Card : PowerColor HD4890 [Xfire]
ATI Catalyst Version : 9.11
OS: Windows 7 Ultimate
DirectX API: DirectX 10
CrossfireX : ON
Screen Resolution : 1980 x 1080
Motion Blur : OFF -- where available
VSynx : OFF -- where available
Below are the results of the stressing software, along with the settings used. I used two synthetic tests Unigine and Furmark and three real life games Crysis Warhead (via the benchmark utility), Resident Evil 5 (via the benchmark utility), and COD:MW2 (real play)
Unigine Dx11 Benchmarking Tool:
*img263.imageshack.us/img263/7383/unigenesetting.png
*img214.imageshack.us/img214/4350/unigene.png
This tool does not use the second CPU core. And quite strongly stresses the CPU. You can see both GPUs are getting almost 100% stressed. At one point 75% CPU usage = 97% GPU usage. Does not seem like a CPU bottle neck.
Furmark:
*img109.imageshack.us/img109/5074/furmarksetting.png
*img214.imageshack.us/img214/4076/furmark.png
This tool also does not understand multi CPU cores. At one point 100% CPU Usage = 99% GPU usage. Could be a CPU bottle neck.
Crysis Warhead:
*img63.imageshack.us/img63/231/crysiswarheadsetting.png
*img63.imageshack.us/img63/497/crysiswarhead.png
Here both cores are definately being used by the game. The point highlighted by me shows a nice ratio of the CPUs working average 75% and both GPUs getting 100% load. Seem like no bottle neck here.
Resident Evil 5:
Setting used: AA=2x; Shadow Detail = high ; Texture Detail = high ; Overall Quality = high
*img263.imageshack.us/img263/5229/residentevilv.png
Here again both cores are being used. The CPU usage here is higher, but not at 100%. The GPUs are not being full used here. Here I feel, the game was being rendered 'easy'. So the GPUs were not stressed 100%. The benchmark showed me around 115 FPS overall.
COD:Modern Warefare 2:
Setting Used: AA=2x;Shadows,Specular Map,Depth of Field, Soft Smoked Edges, Ragdoll, Bullet Impact = Y ; Texture Quality = Automatic
*img63.imageshack.us/img63/944/codmw2.png
Both CPU cores are being used, and the GPUs are well stressed too. The GPUs are getting enough data, without the CPU stopping the process.
Overall Summary:
My system has a CPU bottleneck -- but not major hindrance. It is able to render satisfactory on 1980 x 1080 and stresses the cards well. FPS is decent. Yes I should have mentioned FPS here, but honestly, all games are playable.
So what you guys say...! Hope this helps others figure out their systems, or future purchases/upgrades.
